hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "eric baldeschwieler (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-2560) Combining multiple input blocks into one mapper
Date Wed, 09 Jan 2008 23:28:34 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-2560?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12557477#action_12557477
] 

eric baldeschwieler commented on HADOOP-2560:
---------------------------------------------

Nodes operate at different rates.  Failures happen.  In the face of several jobs running,
some nodes may not even become available in a timely manner.  I think a static approach will
not allow both the performance gains desired and preservation of reasonable throughput.

The current system takes full advantage of mapping jobs to nodes dynamically.  A static combination
of splits will break all of this.  One could perhaps do something like what you suggest dynamically
in the JT when a TT requests a new job.  This might be a good compromise implementation. 
This would also let you observe some global statistics on speed of maps & size of outputs
which would let you optimize cluster sizes.  Of course doing this all dynamically on the TTs
might use fewer JT resources.



> Combining multiple input blocks into one mapper
> -----------------------------------------------
>
>                 Key: HADOOP-2560
>                 URL: https://issues.apache.org/jira/browse/HADOOP-2560
>             Project: Hadoop
>          Issue Type: Bug
>            Reporter: Runping Qi
>
> Currently, an input split contains a consecutive chunk of input file, which by default,
corresponding to a DFS block.
> This may lead to a large number of mapper tasks if the input data is large. This leads
to the following problems:
> 1. Shuffling cost: since the framework has to move M * R map output segments to the nodes
running reducers, 
> larger M means larger shuffling cost.
> 2. High JVM initialization overhead
> 3. Disk fragmentation: larger number of map output files means lower read throughput
for accessing them.
> Ideally, you want to keep the number of mappers to no more than 16 times the number of
 nodes in the cluster.
> To achive that, we can increase the input split size. However, if a split span over more
than one dfs block,
> you lose the data locality scheduling benefits.
> One way to address this problem is to combine multiple input blocks with the same rack
into one split.
> If in average we combine B blocks into one split, then we will reduce the number of mappers
by a factor of B.
> Since all the blocks for one mapper share a rack, thus we can benefit from rack-aware
scheduling.
> Thoughts?

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message